All articles
Analysis

How Do You Perform Call Detail Record Analysis?

Lantern IntelligenceNovember 13, 20267 min read

Call detail record analysis means examining provider-generated communication records to identify contacts, timing, frequency, and investigative leads. Start by validating the return and its field definitions, then normalize numbers and timestamps, compare patterns against other evidence, and document both findings and limitations. A call record alone does not establish who used a device or what was said.

What is a call detail record?

A call detail record, or CDR, is a provider-generated record describing a telecommunications event. Depending on the provider and production, it may contain originating and destination numbers, event time, duration, direction, disposition, and network identifiers.

Some productions also include messaging metadata or data-session records. Those events need separate interpretation: a data session is not a voice call, and a message event does not necessarily include message content.

Working definition: Call detail record analysis is the structured review of telecommunications metadata to test questions about communication activity while preserving the distinction between recorded events and inferred behavior.

Begin with a specific question. “Which numbers contacted this line during the relevant window?” is more testable than “Who is connected to this person?”

What should you check before analyzing call records?

Treat intake as an analytical step, not an administrative formality. A spreadsheet can look complete while omitting a date range, event category, or explanatory attachment.

Use this intake checklist:

  1. Identify the source. Record the provider, production identifier, receipt date, filenames, and supplied documentation.
  2. Confirm coverage. Note the requested account or number, the produced period, and any stated omissions.
  3. Preserve originals. Keep received files unchanged and work from documented copies.
  4. Locate definitions. Find the provider’s field legend, time-zone explanation, disposition codes, and relevant network reference tables.
  5. Check readability. Inspect spreadsheet formatting, PDF tables, scanned pages, and truncated identifiers.
  6. Record uncertainties. Maintain an open-questions list before drawing conclusions.

Hash files using your established evidence-handling process where appropriate. Record transformations separately so another reviewer can reproduce the working dataset without changing the original evidence.

Which CDR fields matter most?

There is no universal CDR layout. Interpret each field using the documentation supplied for that production, rather than assumptions carried over from another provider.

FieldPractical useKey limitation
Originating and destination numbersIdentify recorded communication endpointsAn endpoint is not verified user identity
Event timestampOrder activity and compare recordsTime zone and event-time definition may vary
DurationCompare recorded event lengthsZero or short duration needs disposition context
Direction and dispositionDistinguish incoming, outgoing, attempted, or completed eventsCodes are provider-specific
Device or subscriber identifiersExamine account or equipment associationsAssociation does not prove possession at an event time
Cell-site or sector identifiersExamine recorded network associationsA serving site is not a precise device location

Keep identifiers as text where needed. Spreadsheet software can strip leading zeros, alter long values, or display them in scientific notation. Compare imported values with the source before relying on filters or summaries.

How do you normalize numbers and timestamps safely?

Normalization makes records comparable without erasing their original meaning. Keep original values alongside derived fields and document every conversion.

  1. Create a source reference. Retain the filename, sheet or page, and row identifier for each working record.
  2. Separate event types. Distinguish voice, messaging, data, and other categories before calculating totals.
  3. Standardize numbers cautiously. Use a consistent international format when country context supports it. Preserve short codes, withheld values, and unresolved numbers separately.
  4. Normalize time explicitly. Record the original timestamp, documented source time zone, converted analysis time, and conversion rule.
  5. Check daylight-saving transitions. Flag ambiguous or nonexistent local times rather than silently resolving them.
  6. Review apparent duplicates. Determine whether matching rows represent repeated production, separate call legs, or distinct events.

Do not remove rows simply because their timestamps and numbers match. Network records may represent different parts of one communication. Count raw records and interpreted events separately when the distinction matters.

If the time zone is undocumented, mark it unresolved. A precise-looking timestamp can still carry substantial uncertainty.

How do you identify useful communication patterns?

Start with descriptive summaries, then test narrow hypotheses. Useful views include event counts by counterpart, activity by day, incoming versus outgoing events, and communications within a defined window.

For example, suppose a fictional review finds repeated calls between two numbers shortly before several documented transactions. That pattern supports closer examination of timing and account attribution. It does not, by itself, establish coordination or knowledge of the transactions.

Use a repeatable sequence:

  1. Define the relevant event window and explain why you selected it.
  2. Establish a comparison period where records permit.
  3. Identify recurring contacts and changes in frequency or duration.
  4. Inspect the underlying rows behind each apparent pattern.
  5. Compare findings with independent evidence.
  6. Document plausible alternative explanations.

High call volume can reflect customer service, work scheduling, automated systems, or ordinary personal contact. Low volume does not prove a weak relationship; communications may occur through services absent from the production.

For a broader method of evaluating connections, see Link Analysis Law Enforcement Teams Can Use to Map Connections.

What can call records tell you about identity and location?

Separate three propositions: a number appears in a record, an account is associated with that number, and a particular person used it during an event. Each requires different support.

Subscriber information may help establish an account association. A phone extraction, relevant message, interview, or other evidence may strengthen or challenge user attribution. Shared devices, reassigned numbers, forwarding, and spoofing can complicate interpretation, depending on the records and event type.

Location requires similar restraint. A cell-site identifier describes a recorded network association, not an exact handset position. Coverage depends on network configuration, terrain, buildings, and operating conditions.

Do not turn a serving-site entry into an exact location claim. Historical site information and specialist interpretation may be necessary. Likewise, an absence of records does not prove an absence of communication or activity.

How can Lantern support call detail record analysis?

Lantern Intelligence is an AI evidence platform for investigators. It accepts any evidence file, including warrant returns from any provider, phone extractions, PDFs, scans, photos, video, spreadsheets, email, bank records, and call records. OCR turns scans, images, and handwriting into searchable text.

Use Upload Case to bring evidence together, then review material through live areas including Files, Records, Search, Persons, and Messages. The “Ask the evidence” AI chat can help you explore questions, while AI Insights provides an intel brief and next steps. Verify AI-generated statements against source records before using them as findings.

For scanned tables, compare extracted text with the page image, especially telephone numbers, timestamps, and column boundaries. Searchability does not eliminate transcription risk. Timeline and Relationship graph are coming soon, not live analysis tools.

Lantern runs in private AWS or on-prem on Google Gemma 4; evidence is never sent to an outside AI service. CJIS, FedRAMP, and SOC 2 are coming soon; they are not current certifications.

See how Lantern works, explore the use cases, or review pricing. Free sign-up includes 10 credits.

What should your final analytical note include?

A useful note makes the reasoning inspectable. Separate observations, interpretations, and unresolved questions rather than blending them into one narrative.

Include this closing checklist:

  1. Scope: Numbers, files, event types, and dates reviewed.
  2. Method: Field definitions, time conversions, filters, and duplicate-handling decisions.
  3. Findings: Specific observations with source references and reproducible counts.
  4. Corroboration: Evidence that supports or conflicts with each interpretation.
  5. Limits: Missing records, uncertain attribution, unresolved timestamps, and location caveats.
  6. Next steps: Targeted questions that would confirm or challenge the working explanation.

The goal is not the largest contact list. It is a transparent account of what the records show, what remains uncertain, and which questions deserve further review.

Start free — get 10 credits.

Put this to work on your own case

Sign up free and get 10 credits — enough to process a typical warrant return and see the messages, people, AI summary and recommended next steps.

call detail record analysiscall detail recordsCDR analysisphone records investigationtelecommunications evidencecall record review
How it worksStart free